Engine control systems utilize adaptive learning to correct sensor readings, or compensate for component wear. In particular, exhaust gas sensors coupled to an engine exhaust are typically used for such adaptive learning. For example, the sensor reading can be used to correct for air-flow measurement errors, and changes in fuel injectors due to aging.
The inventors herein have developed an engine control methodology that allows efficient engine operation with some of the cylinders inducting air with no injected fuel. The inventors have also found that when using such a system with conventional adaptive learning methods, degraded results are obtained. For example, when the system includes exhaust gas sensors exposed to air from cylinders without injected fuel, there is no information as to the amount of fuel injected into other cylinders, and thus no information to be had from that adaptation algorithm. As such, when the algorithm attempts to incorporate this sensor data, degraded results are obtained.